https://github.com/ambarfield/Barfield_Rathod_Fitzgerald_ENV872_EDA_FinalProject.git
This project intends to discover whether the diet of polar bears has changed over time in Alaska. Polar bears (Ursus maritimus) are apex predators in the Arctic ecosystem. Polar bears depend on sea ice for hunting, typically to use a platform to hunt seals. As global temperatures rise, melting sea ice has led to habitat loss for polar bears, forcing them onto land. Without access to ice, researchers have found that polar bears enter a state of fasting, increasing the time-period during which they must rely on their fat reserves (Pilfold et al., 2016). Within this context, it is critical to understand whether the diets of polar bears are evolving, if at all, in response to climate change. This will allow researchers to predict how Polar bear populations will change based on sea ice loss predictions and whether polar bears are adapting alternative feeding strategies (e.g. terrestrial creatures) to improve survival.
We chose a USGS dataset that includes carbon and nitrogen isotope concentrations measured in polar bear hair between 1983-2017 in the Beaufort and Chukchi Seas. Polar bears were located by helicopter and immobilized with a dart containing zolazepam-tiletamine, a common anaesthetic used in veterinary medicine. Hair was sampled by pulling 10-20 hairs from one of the upper forearms of polar bears. This dataset was chosen because elements found in hair (Nitrogen, Carbon) indicate the macronutrient composition of diets (protein vs. fat) (Rode et al., 2023).
Primary Question: Have polar bear diets change over time in the Beaufort and Chukchi Sea?
The dataset used in this analysis was obtained from the United States Geological Survey (USGS) website, more specifically from the Alaska Science Center (ACS). It catalogues the carbon and nitrogen isotope concentrations found in polar bear hair samples in the Chukchi and Beaufort Sea. The period for obtaining these samples are as follows: Chukchi Sea samples (Russia/Alaska) were collected from periods of 1987-1989 and 2008-2017 and Beaufort Sea samples (Alaska) were collected from periods of 1983-1989 and 2004-2016. The basic data structure of this dataset is outlined below:
| Column Name | Description |
|---|---|
| BearID | Unique numerical identifier for each bear |
| Sex | The sex of the bear determined at capture |
| Age | The age of bear determined through teeth or best estimate |
| Capture.Date | The date bear was captured and hair was collected |
| Year | The year bear was captured and sampled |
| Pop | Subpopulation boundary where bear was captured |
| TypicalSummerHabitat | Whether bear summered on ice or land (only for Chukchi Sea bears) |
| SummerHabitatDuringHairGrowth | Whether bear summered on ice or land based on sample |
| Lat | Latitude of capture |
| Long | Longitude of capture |
| d15N.Air | Stable nitrogen (15N/14N) isotope ratios in δ notation, as parts per thousand (‰) deviation from the primary standard, atmospheric air |
| d13C.VPDB | Stable carbon (13C/12C) isotope ratios in δ notation, as parts per thousand (‰) deviation from the primary standards Vienna Pee Dee Belemnite |
| Percent.N | The percent of nitrogen by mass measured in the hair |
| Percent.C | The percent carbon by mass measured in the hair |
Data was wrangled for most of the figures by first taking out two unnecessary columns: “TypicalSummerHabitat” and “SummerHabitatDuringHairGrowth”. These columns only had data for about half of the recorded hair samples; however, they were analyzed through a linear regression. All other samples with “NA” fields were taken out of the cleaned dataset before the bulk of the data anlysis. For the time series portion of the analysis, data was wrangled by changing the Capture.Date column from a factor into a date, and then organizing the date column in ascending order. Any year that contained sampling dates that fell outside the months of March, April, and May were then excluded from the analysis. The dataset was also re-organized by dividing the dataset into early years (1983-1989) and late years(2004-2017) due to there being an absence of data from the years of 1990-2003, which needed to be excluded for the time series analysis.
| Count, # | |
|---|---|
| F | 357 |
| M | 283 |
| Stable Nitrogen Isotope Ratio, ppm | Stable Carbon Isotope Ratio, ppm | Percent Nitrogen, % | Percent Carbon, % | |
|---|---|---|---|---|
| Mean Value | 20.74 | -16.09 | 15.24 | 44.14 |
In order to explore how this dataset was represented, it was important to see how it was distributed spatially throughout the Beaufort and Chukchi Sea. The map below shows the distribution of polar bear samples by the year they were collected. This gives viewers an idea of the relative spatial distribution of the data over time.
The map below shows the relative spatial distribution of all polar bear hair samples separated out by sex.
Analysis Linear Regressions - Correlating Summer Habitats and Nitrogen Isotopes
##
## Call:
## lm(formula = d15N.Air ~ Habitat, data = polarbearhair_hab_clean)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.24839 -0.33710 0.06667 0.48790 1.66667
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 20.1484 0.1188 169.54 < 2e-16 ***
## Habitat 0.8849 0.1544 5.73 2.04e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.6617 on 74 degrees of freedom
## Multiple R-squared: 0.3073, Adjusted R-squared: 0.298
## F-statistic: 32.83 on 1 and 74 DF, p-value: 2.039e-07
Through this linear regression model, we wanted to test whether there is any effect of polar bear typical habitat (on ice vs. land) on the polar bear diet (via the ratio of stable nitrogen isotopes). The output of this regression suggests that there is a statistically significant relationship between habitat and diet (p-value of 2.04E-07 < 0.01). Given that our dataset had a relatively small sample of habitat data, this analysis should be repeated with a larger dataset.
## `geom_smooth()` using formula = 'y ~ x'
Figure 1. The trend in Figure 1 suggests that the stable nitrogen isotopes across polar bears have remained more or less consistent over the ~15 year period.
## `geom_smooth()` using formula = 'y ~ x'
Figure 2. The trend in Figure 2 suggests that the percentage of nitrogen found in polar bear hair samples has remained more or less consistent over the ~15 year period. Since these nitrogen variables represents the trophic level of polar bear’s prey, our results indicate that polar bears have not changed the species they feed on.
## `geom_smooth()` using formula = 'y ~ x'
Figure 3. The trend is Figure 3 suggests that the stable carbon isotopes across polar bears have remained more or less consistent over the ~15 year period.
## `geom_smooth()` using formula = 'y ~ x'
Figure 4. The trend is Figure 4 suggests that the percentage of carbon found in polar bear hair samples has remained more or less consistent over the ~15 year period. Since these carbon variables represents whether polar bears feed on marine or terrestrial sources, our results indicate that polar bears have not changed their feeding patterns.
## [1] FALSE
## [1] FALSE
Figure 5. The graph above depicts the nitrogen and carbon stable isotope trends from polar bear hair samples collected in the Beaufort and Chukchi Seas between 1983 - 1989.
Figure 6. The graph above depicts the nitrogen and carbon stable isotope trends from polar bear hair samples collected in the Beaufort and Chukchi Seas between 2004-2017.
Figure 7. The graph above depicts the decomposition time series for carbon stable isotope concentrations collected from polar bears between 2004-2017 in the Beafort and Chukchi Seas.
Figure 8. The graph above depicts the decomposition time series for nitrogen stable isotope concentrations collected from polar bears between 2004-2017 in the Beafort and Chukchi Seas.
Figure 9. The graph above depicts the trend data for stable carbon isoptope concentrations from polar bear samples collected in the Beafuort and Chukchi Seas from 2004-2017.
Figure 10. The graph above depicts the trend data for stable nitrogen isoptope concentrations from polar bear samples collected in the Beafuort and Chukchi Seas from 2004-2017.
#Remove seasonality from data
PolarBear.Late.Nitrogen.NoSeasonality <- PolarBearLate.Nitrogen.ts - PolarBearLate.Nitrogen.Decomposed$time.series[, "seasonal"]
PolarBear.Late.Carbon.NoSeasonality <- PolarBearLate.Carbon.ts -
PolarBearLate.Carbon.Decomposed$time.series[, "seasonal"]
#Perform Man-Kendall Trend Test
PolarBear.Late.Nitrogen.Trend <- Kendall::MannKendall(PolarBear.Late.Nitrogen.NoSeasonality)
PolarBear.Late.Carbon.Trend <- Kendall::MannKendall(PolarBear.Late.Carbon.NoSeasonality)
#Summarize Trend Tests
summary(PolarBear.Late.Nitrogen.Trend)
## Score = -4857 , Var(Score) = 10779577
## denominator = 105048
## tau = -0.0462, 2-sided pvalue =0.13913
summary(PolarBear.Late.Carbon.Trend)
## Score = 3450 , Var(Score) = 10779642
## denominator = 105075.5
## tau = 0.0328, 2-sided pvalue =0.29349
Our results indicate that polar bear diets have largely not changed during the time period (2004-2017). We did find that there is a statistically significant relationships between the stable nitrogen isotopes (which indicate the trophic level of the prey that polar bears feed on) and typical polar bear habitats (on ice vs. land). This is likely because polar bears that spend more time on ice feed on seals (higher trophic level). In compared to polar bears that spend more time on land feed on smaller mammals like rodents or ox (lower trophic level). However, it is important to note that the sample size for the linear regression was smaller than the rest of the dataset (N= 106 vs. N=640). To confirm if this relationship be extended to a broader population, this analysis should be repeated with a larger sample size.
The results of the performed Man-Kendall trend tests reveal no significant trends in both carbon and nitrogen isotope concentrations collected from the polar bear hair samples between 2004-2017. The nitrogen and phosphorous concentrations had tau values of -0.0462 and 0.0328, respectively, but both p-values were greater than 0.05, indicating no statistically significant trends.
As a next step, this data can be connected to datasets on prey sample (also collected by USGS) to identify what polar bears feed on specifically. In addition, data on polar bear weights can also be collected over time as this would indicate if polar bears are fasting for longer periods every year.
Pilfold, N. W., Hedman, D., Stirling, I., Derocher, A. E., Lunn, N. J., & Richardson, E. (2016). Mass Loss Rates of Fasting Polar Bears. Physiological and Biochemical Zoology: PBZ, 89(5), 377–388. https://doi.org/10.1086/687988
Rode, K.D., 2021, Carbon and nitrogen isotope concentrations in polar bear hair and prey from the Alaska Beaufort and Chukchi Seas, 1978-2019: U.S. Geological Survey data release, https://doi.org/10.5066/P9KM5FT2
Rode, K. D., Taras, B. D., Stricker, C. A., Atwood, T. C., Boucher, N. P., Durner, G. M., Derocher, A. E., Richardson, E. S., Cherry, S. G., Quakenbush, L., Horstmann, L., & Bromaghin, J. F. (2023). Diet energy density estimated from isotopes in predator hair associated with survival, habitat, and population dynamics. Ecological Applications, 33(2), e2751. https://doi.org/10.1002/eap.2751